Dynamic Item Pick Pose Recovery for Robotic Placement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing robotic systems face challenges in efficiently determining the pose of items during pick and place operations, leading to delays and inefficiencies in automated tasks, particularly in repositioning items among containers.
Innovation Solution
A computer vision-based approach using fiducial markers, cameras, and a computing environment to estimate the pose of items through point cloud processing, allowing for real-time autonomous path planning and precise placement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional robotic systems are used for item handling, then automation is achieved, but pose determination accuracy and speed are insufficient
Solution Approach 1:
The patent replaces traditional mechanical sensor systems with a computer vision-based system using cameras and fiducial markers to determine item pose. The vision system processes images through point cloud generation and pose estimation algorithms to rapidly and accurately determine item orientation and position, achieving both high precision and speed in pose determination.
Solution Approach 2:
The patent uses fiducial markers as visual copies or representations of the item's pose information. These markers provide known geometric patterns that can be easily detected and processed by vision algorithms, serving as informative copies that enable accurate pose estimation without requiring direct measurement of the item itself.
2Manufacturing precision
If pose determination is performed in real-time during item movement, then placement precision is improved, but system complexity increases
Solution Approach 1:
The patent performs preliminary actions by pre-defining the geometric patterns of fiducial markers and pre-establishing the relationship between marker detection and pose calculation. This preparation allows the real-time pose determination to be more efficient, as the system only needs to detect pre-known patterns rather than perform complex analysis during item movement.
Solution Approach 2:
The patent segments the pose determination process into distinct functional components: image capture, fiducial marker detection, point cloud generation, and pose estimation. This segmentation allows each component to be optimized independently and simplifies the overall system architecture by dividing the complex task into manageable modules.
Data Source
AI summary
Dynamic pick pose recovery is described. The concepts can be relied upon to determine the pose of an item during a process for picking, moving, and placing the item among totes, or other uses. In one example, an item can be picked using a robotic arm, and the item can be repositioned for placement at a second location. Images of the item can be captured as the item moves. The images can be segmented to determine masks of the item. The masks can be projected into a point cloud within a bounding volume, to generate contours. Points of the point cloud that are positioned outside the contours can be discarded, to find a subset of points representative of a pose of the item. The position and orientation of the item can be calculated, for placement of the item at a second location, based on the pose of the item.